基于模糊连接度的交互式活动轮廓模型

Interactive active contour model based on fuzzy connectedness

由于活动轮廓模型能量函数是非凸的,图像分割的结果易于陷入局部最优。为了克服该问题,提出一种基于凸活动轮廓模型的交互式彩色图像分割方法。该方法的新能量函数不仅充分利用边缘信息和颜色信息,还包含一种新定义的空间位置信息。通过模糊连接度构造空间位置信息,将其自适应地融合到活动轮廓模型中。在数值优化过程中,采用分裂Bregman方法获得新模型的全局最优解。针对多幅彩色自然图像作对比实验,结果表明新方法能够准确、快速地得到理想的分割结果。

The energy function of active contour model is not convex, so the solution of a minimization problem is prone to fall into a local minimum. Dealing with the problem, this paper proposed a novel convex active contour model for interactive color image segmentation. It defined a new energy function including not only image edge and color information, but also a spatial information term which was adaptively fused with the color information. It defined the spatial information term by fuzzy con- nectedness and introduced to get more accurate segmentation results. In the energy numerical minimization process, it used split Bregman method to obtain a global solution. Experimental results demonstrate that the proposed method can accurately and quickly get desired segmentation results.